Google to invest up to $40B in Anthropic in cash and compute

```json { "title": "Google to Invest Up to $40B in Anthropic Amid Mythos AI Surge", "metaDescription": "Google commits $10B to Anthropic now, with $30B more tied to performance targets, as the AI startup's Mythos cybersecurity model reshapes the industry.", "content": "<h2>Google Commits Up to $40 Billion to Anthropic in Landmark AI Deal</h2>\n\n<p>Google announced on April 24, 2026, that it will invest up to $40 billion in Anthropic PBC, the AI safety company behind the Claude family of models. According to Bloomberg, the deal structures $10 billion as an immediate cash investment, with up to $30 billion more contingent on Anthropic hitting certain performance targets. The announcement arrives just days after Amazon disclosed a similarly structured deal worth up to $25 billion, cementing Anthropic as one of the most heavily capitalized AI companies in the world — and underscoring an intensifying race among big tech players to secure both equity stakes and computing capacity in frontier AI.</p>\n\n<p>The initial $10 billion tranche values Anthropic at $350 billion, matching the valuation from a funding round that closed in February 2026. That earlier round raised $30 billion and valued Anthropic at $380 billion post-money, according to reporting by the Cyprus Mail. The new Google deal does not appear to revise that valuation upward, but the scale of committed capital — from multiple directions simultaneously — signals that investor confidence in Anthropic's commercial trajectory remains exceptionally high.</p>\n\n<h2>What the Deal Actually Includes: Cash, Compute, and a Three-Way Chip Agreement</h2>\n\n<p>Beyond the headline dollar figure, the Google-Anthropic arrangement includes a substantial computing component. Google Cloud has agreed to deliver five gigawatts of computing power to Anthropic across a five-year window, with the potential for additional gigawatts thereafter, according to Bloomberg as reported by Yahoo Finance. That level of compute access is not incidental — it is central to Anthropic's ability to train and operate increasingly powerful models at scale.</p>\n\n<p>The computing arrangement also builds on a three-way agreement between Anthropic, Google, and Broadcom that was announced earlier in April 2026. Broadcom, which designs custom AI chips for major cloud providers, is now part of the infrastructure architecture underpinning Anthropic's model development pipeline. This multi-party structure reflects a broader industry pattern in which AI labs are not simply shopping for cloud credits but are negotiating deeply integrated hardware and infrastructure partnerships to future-proof their capacity.</p>\n\n<p>Prior to the April 24 announcement, Google had already invested more than $3 billion in Anthropic and reportedly held approximately a 14% stake in the company, according to CNBC. Google's relationship with Anthropic dates to 2023, when it made an initial $300 million investment for roughly 10% of the company, followed by a further $2 billion investment months later. The new deal substantially deepens that relationship, though the exact post-deal ownership breakdown was not disclosed in available reporting.</p>\n\n<h2>Surging Revenue and the Claude Code Effect</h2>\n\n<p>The investment frenzy is being driven in large part by Anthropic's remarkable revenue growth. According to Bloomberg, as reported by BNN Bloomberg, Anthropic's annual run-rate revenue surpassed $30 billion in April 2026 — up from approximately $9 billion at the end of 2025. That represents more than a threefold increase in roughly four months, a growth rate that is difficult to contextualize against conventional software benchmarks.</p>\n\n<p>Bloomberg and Yahoo Finance attribute a significant share of that growth to booming demand for Claude Code, Anthropic's AI agent built to accelerate software development. Claude Code positions Anthropic squarely in the AI coding assistant market, where enterprises are increasingly willing to pay for tools that reduce development cycle times. The velocity of revenue growth suggests that adoption has moved well beyond early pilots into production-scale deployments at large organizations.</p>\n\n<p>Amazon's deal, disclosed just days before Google's announcement, offers up to $5 billion now with an option to invest up to $20 billion more tied to commercial milestones, according to Reuters and BNN Bloomberg. Both deals value Anthropic at $350 billion. The near-simultaneous nature of the two announcements — from two of the largest cloud providers in the world — points to a competitive dynamic in which neither Amazon nor Google can afford to be the party that does not have a deep, structural relationship with one of the leading frontier AI labs.</p>\n\n<h2>Claude Mythos Preview: The Cybersecurity Model Too Powerful to Release Publicly</h2>\n\n<p>Running parallel to the investment news is a separate but closely related development: the limited release of Claude Mythos Preview, a new AI model that Anthropic describes as <em>"strikingly capable at computer security tasks"</em> on its frontier red team blog. Announced on April 7, 2026, as part of Project Glasswing, Mythos Preview has not been made generally available — and Anthropic has stated it does not plan to do so.</p>\n\n<p>The model's capabilities, as documented by Anthropic's own red team, are striking. According to the Anthropic frontier red team blog, Mythos Preview developed working exploits in a Firefox JavaScript engine vulnerability 181 times, compared to Claude Opus 4.6, which succeeded only two times out of several hundred attempts on the same task. According to Fortune, Mythos Preview demonstrated an 83 percent success rate in exploit creation on the first attempt. According to Anthropic's official Project Glasswing page, the model has been used to identify thousands of zero-day vulnerabilities across every major operating system and every major web browser — with over 99% of those vulnerabilities still unpatched at the time of publication, preventing full public disclosure.</p>\n\n<p>An Anthropic spokesperson described the model's development goals plainly: <em>"We're developing a general purpose model with meaningful advances in reasoning, coding, and cybersecurity."</em> Logan Graham, head of Anthropic's frontier red team, offered a more pointed characterization, saying Mythos Preview is <em>"extremely autonomous"</em> and has <em>"sophisticated reasoning capabilities that give it the skills of an advanced security researcher."</em></p>\n\n<p>Project Glasswing currently grants access to 12 formal partner organizations and approximately 40 organizations in total, according to TechCrunch. Named partners include Amazon, Apple, Broadcom, Cisco, CrowdStrike, the Linux Foundation, Microsoft, and Palo Alto Networks. Anthropic has committed up to $100 million in Mythos Preview usage credits and $4 million in direct donations to open-source security organizations as part of the initiative.</p>\n\n<h2>Regulatory Gaps and an Unauthorized Access Incident</h2>\n\n<p>The controlled release has not been without controversy. According to Axios, the U.S. Cybersecurity and Infrastructure Security Agency — the country's primary civilian cyber defense body — does not have access to Mythos Preview, even as other government agencies including the NSA are using it. That gap has drawn attention in Washington, where policymakers are grappling with how to assess and govern AI tools capable of autonomous vulnerability discovery and exploit generation at scale.</p>\n\n<p>Complicating the picture further, a private online forum reportedly gained unauthorized access to Mythos Preview through a third-party vendor, according to Euronews Next on April 22, 2026. Anthropic confirmed it was looking into the matter, with a spokesperson stating: <em>"We're investigating a report claiming unauthorised access to Claude Mythos Preview through one of our third-party vendor environments."</em> Anthropic added that there was no evidence its own systems were impacted.</p>\n\n<p>The incident highlights a fundamental tension in deploying dual-use AI capabilities through third-party commercial channels: even a tightly controlled release depends on the security posture of every vendor in the access chain. For a model that Anthropic itself considers too dangerous for public release, a single compromised vendor environment is a significant exposure.</p>\n\n<p>Despite these concerns, Anthropic's position on the potential upside of Mythos for defense is clear. An unnamed Anthropic official told Axios: <em>"There's an opportunity here to give a shot in the arm to defense and to keep pace with this long-standing trend where offense exploitation had an advantage."</em></p>\n\n<h2>Why This Matters: AI Infrastructure, Security, and the Race for Compute</h2>\n\n<p>Taken together, the Google investment and the Mythos developments paint a picture of an AI industry in which the stakes — financial, geopolitical, and security-related — have escalated dramatically in a short period of time. Anthropic's revenue tripling in four months, two of the world's largest cloud providers committing a combined potential of $65 billion, and a cybersecurity model capable of autonomously discovering zero-day vulnerabilities across every major platform: these are not incremental developments. They represent a qualitative shift in what frontier AI labs are building and how seriously major institutions — corporate and governmental — are treating the implications.</p>\n\n<p>The five-gigawatt compute commitment from Google Cloud alone reflects how thoroughly the economics of frontier AI have become inseparable from energy and infrastructure policy. The compute demands of training and deploying models like Mythos are not measured in server racks — they are measured in power plants. That reality is reshaping how AI companies negotiate, who they partner with, and what kinds of multi-party agreements are required to sustain model development at the frontier.</p>\n\n<p>For enterprises and professionals tracking these developments, the practical takeaway is that AI capabilities in coding, security research, and autonomous task execution are advancing faster than most governance frameworks can accommodate — and that the organizations with early, structured access to tools like Mythos will have significant advantages in both offense and defense contexts.</p>\n\n<p>For more tech news, visit our <a href=\"/news\">news section</a>.</p>\n\n<h2>What to Watch Next</h2>\n\n<p>Several questions remain open as of April 24, 2026. Anthropic has not disclosed how or whether CISA will eventually gain access to Mythos Preview, nor has it clarified the outcome of its investigation into the third-party vendor breach. The $30 billion contingent tranche of Google's investment is tied to performance targets that have not been publicly defined, leaving the ultimate size of Google's stake in Anthropic uncertain. And as Anthropic's revenue run-rate climbs, the question of if and when the company pursues a public offering — having last been valued at $380 billion post-money — will attract increasing scrutiny from financial markets. The Mythos model's long-term deployment strategy, including whether a safer or more restricted version will eventually reach broader audiences, also remains unresolved.", "excerpt": "Google announced on April 24, 2026, a commitment to invest up to $40 billion in Anthropic, with $10 billion in cash upfront and $30 billion more tied to performance targets, alongside a five-gigawatt computing deal through Google Cloud. The announcement follows a similar multi-billion-dollar deal from Amazon and comes as Anthropic's annual run-rate revenue surpassed $30 billion — more than tripling since the end of 2025. Meanwhile, Anthropic's Claude Mythos Preview, a cybersecurity-focused AI model deemed too powerful for public release, has drawn regulatory scrutiny and was reportedly accessed without authorization through a third-party vendor breach.", "keywords": ["Google Anthropic investment", "Anthropic $40 billion", "Claude Mythos Preview", "Project Glasswing", "AI cybersecurity model"], "slug": "google-invest-40-billion-anthropic-mythos-ai-2026" } ```